M-Predictive Spliner: Enabling Spatiotemporal Multi-Opponent Overtaking for Autonomous Racing

Abstract

Unrestricted multi-agent racing presents a significant research challenge, requiring decision-making at the limits of a robot's operational capabilities. While previous approaches have either ignored spatiotemporal information in the decision-making process or been restricted to single-opponent scenarios, this work enables arbitrary multi-opponent head-to-head racing while considering the opponents' future intent. The proposed method employs a Kalman Filter (KF)-based multi-opponent tracker to effectively perform opponent Re-Identification (reID) by associating them across observations. Simultaneously, spatial and velocity Gaussian Process Regression (GPR) is performed on all observed opponent trajectories, providing predictive information to compute the overtaking maneuvers. This approach has been experimentally validated on a physical 1:10 scale autonomous racing car achieving an overtaking success rate of up to 91.65% and demonstrating an average 10.13%-point improvement in safety at the same speed as the previous State-of-the-Art (SotA). These results highlight its potential for high-performance autonomous racing.